[HADOOP-15297] Make S3A etag => checksum feature optional

• Type: Sub-task
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 3.1.0
• Fix Version/s: 3.1.0
• Component/s: fs/s3
• Target Version/s: 3.1.0
• Assignee: Steve Loughran
• Reporter: Steve Loughran
• Created: 07/Mar/18 20:48
• Updated: 12/Mar/18 14:23
• Resolved: 12/Mar/18 14:03

Description

HADOOP-15273 shows how distcp doesn't handle non-HDFS filesystems with checksums.

Exposing Etags as checksums, HADOOP-13282, breaks workflows which back up to s3a.

Rather than revert I want to make it an option, off by default. Once we are happy with distcp in future, we can turn it on.

Why an option? Because it lines up for a successor to distcp which saves src and dest checksums to a file and can then verify whether or not files have really changed. Currently distcp relies on dest checksum algorithm being the same as the src for incremental updates, but if either of the stores don't serve checksums, silently downgrades to not checking.

Issue Links

• is caused by: HADOOP-13282 S3 blob etags to be made visible in S3A status/getFileChecksum() calls
• is depended upon by: HADOOP-15273 distcp can't handle remote stores with different checksum algorithms